Browser demo

https://scpi-sim.steitec.net is a small test bench you can drive from your browser: a Rohde & Schwarz HMP4040 power supply whose outputs feed two loads and a small board, a Keysight DAQ970A that measures the loads, and a Rigol DS1054Z oscilloscope. All three are scpi-sim instruments with their profiles — the same code that pip install scpi-sim gives you, running in the page.

What you see

Area	What it shows
HMP4040 front (top left)	the four channels: actual voltage and current, the setpoints, CV/CC/OFF, the channel LED and the master OUTPUT button. ERR lights up while the error queue is not empty.
Wiring line	CH1 → 24 Ω ← DAQ 101 and CH2 → 10 Ω ← DAQ 102: channel 1 drives a 24 Ω load that DAQ channel 101 measures, channel 2 a 10 Ω load on channel 102.
DAQ970A front (bottom left)	the display of the built-in multimeter (the last reading in VDC or Ω, OVLD when out of range) and the ten relays of the DAQM901A card; the closed relay is highlighted. A yellow badge shows an active fault.
DS1054Z front (below it)	the oscilloscope's screen, drawn from the simulator's own samples and redrawn while the scope runs: CH1 sees a 1 kHz sine, CH2 the 12 V rail. The bands above and below the eight divisions are the headroom a real screen hides but the data has. The badge shows T'D, AUTO, WAIT or STOP, the line below the scales and the trigger. The arrow on the left marks the trigger level, the one on top the trigger moment.
Terminal (right)	pick psu, daq or scope and type any SCPI command. Commands are blue, answers white, comments (#) grey, timeouts orange. ↑/↓ recall earlier commands and keep a half-typed line. The terminal echoes each line as the instrument got it (at most 256 characters, no control characters).
Try these in order, Then the oscilloscope	twelve buttons that type commands for you (see below).
Break something	six faults and a reset (see below).

Every value on the fronts is read from the simulator after each command; the page itself computes nothing.

The guided tour

#	Button	Sent	What happens
1	Who are you?	psu> *IDN?	Rohde&Schwarz,HMP4040,3629.6776.04/100001,2.70 — the identity of the real model
2	Channel 1 to 12 V, output on	psu> INST OUT1; VOLT 12; OUTP:SEL ON;:OUTP:GEN ON	no answer (it is not a query); CH1 lights up, CV, 12 V. The HMP needs both the channel (OUTP:SEL) and the master switch (OUTP:GEN).
3	Measure the current	psu> MEAS:CURR?	0.500 — 12 V into 24 Ω, in the HMP's number format
4	DAQ measures the 12 V rail	daq> MEAS:VOLT:DC? (@101)	about +1.19996E+01 — the DAQ closes relay 101 and measures; small noise like a real meter
5	Drop to 9 V and measure again	psu> VOLT 9, daq> MEAS:VOLT:DC? (@101)	the DAQ now reads about 9 V: the instruments are wired together, so measurements follow what the script switched
6	Send garbage, read the error queue	psu> FOO, psu> SYST:ERR?	-113,"Undefined header"; ERR lights up until the queue is read

And the oscilloscope (Oscilloscope has the whole story):

#	Button	Sent	What happens
7	Measure the sine	scope> :MEAS:ITEM? FREQ,CHAN1, :MEAS:ITEM? VPP,CHAN1	about 1 kHz and 2 V, computed from the samples
8	Read the waveform, a binary block	scope> :WAV:PRE?, :WAV:DATA?	the scaling, then #9000001200<1200 bytes of binary data>: the terminal shows the block by its size
9	Too sensitive: it clips, Vpp reads wrong	scope> :CHAN1:SCAL 0.1, :MEAS:ITEM? VPP,CHAN1	the tops are cut off and Vpp reads about 1 V instead of 2 V, with no error anywhere
10	Back to 500 mV/div, 12 V rail on CH2	scope> :CHAN1:SCAL 0.5, :CHAN2:SCAL 5, :CHAN2:DISP ON, psu> INST OUT1; VOLT 12; ...	the supply's rail appears as a second trace: the instruments are wired together
11	Trigger above the signal: the query hangs	scope> :TRIG:EDG:LEV 1.5, :SING, *OPC?	the badge shows WAIT, and after 2 s the terminal shows a timeout
12	Lower the level: it fires, run again	scope> :TRIG:EDG:LEV 0, *OPC?, :RUN	1, and the scope runs again
13	Catch a board's power-on: inrush on CH3, 3.3 V rail on CH4	psu> INST OUT3; ...; OUTP:SEL OFF, scope> ..., :TRIG:EDG:SOUR CHAN3;:TRIG:EDG:LEV 0.02;:SING, psu> OUTP:SEL ON;:OUTP:GEN ON	the scope stops with one capture: the inrush peak, a load step when the CPU starts, and the rail coming up 2 ms after power-on
14	Read the inrush peak (100 mV/A) and the rail	scope> :MEAS:ITEM? VMAX,CHAN3, :MEAS:ITEM? VMAX,CHAN4	about 0.22 V, so 2.2 A, and about 3.3 V

Typing works too. A few things to try:

• psu> APPL 5,1 after INST OUT2, then OUTP:SEL ON — channel 2 into 10 Ω draws 0.5 A.
• psu> CURR 0.2 on channel 1 at 12 V: the load wants 0.5 A, so the channel goes into current limit (CC) and the voltage drops to 4.8 V.
• daq> MEAS:VOLT:DC? (@101,102) — both rails in one scan.
• psu> VOLT 40, then psu> SYST:ERR? — 40 V is out of range for this model: -222,"Data out of range", and the setpoint stays as it was.
• scope> :TRIG:EDG:SLOP NEG and watch the sine start on its falling edge.
• scope> :TIM:SCAL 0.000001, then scope> :MEAS:ITEM? FREQ,CHAN1: no complete period on the screen, so 9.900000e+37, the manual's "no result".
• scope> :STOP, then scope> :CHAN1:SCAL 2: the capture on a stopped scope stays as it was.
• psu> meas:curr? works, psu> MEASU:CURR? does not — keywords match in short or long form only, as on real instruments.

Break something

These are scpi-sim's fault injection, the feature you use in tests to check that your code copes with misbehaving instruments.

Button	Fault	What your test code would see
DAQ stops answering	no_response, once	the next query gets no answer; after 2 s the terminal shows timeout – no response (VI_ERROR_TMO in PyVISA). The command itself still ran.
DAQ reads +0.5 V off	offset 0.5 V	every reading is 0.5 V high — a drifting meter or a bad calibration. A limit check in your test should catch it.
PSU reports error −222	error -222	an entry appears in the error queue (ERR lights up) without a command causing it — does your code check SYST:ERR??
Scope never triggers	no_trigger	Auto sweep runs unsynchronised, the trace wanders and the badge shows AUTO. A :SING would wait for ever.
Noisy signal	noise 0.1 V	a thick trace, and Vpp reads high
Probe falls off CH1	signal, 0 V on channel 1	a flat line and no frequency any more: does your test notice?
Reset bench	—	clears the faults and puts the instruments back to their power-on state

In your own tests the same faults come from Python (bench.daq.inject("offset", value=0.5)), from pytest, or from the command line (scpi-sim fault daq offset value=0.5). Recipes: Fault injection recipes.

How it works

The site is static — HTML, CSS, a little JavaScript, and Pyodide (CPython compiled to WebAssembly). When the page loads, it

1. starts Pyodide in the browser,
2. installs a scpi-sim wheel built from the same release as the PyPI package,
3. loads the bench from examples/profiles/bench.yaml,
4. sends every command you type to the instrument's process() method — the same method the TCP, VXI-11 and HiSLIP servers call — and draws the fronts from the returned state.

There are no sockets in the browser, so the demo does not start servers; the messages go straight to the instrument models. The oscilloscope's screen reads the latest capture's samples directly, without changing its waveform settings; only what the demo draws differs from what :WAV:DATA? returns (600 points per trace). That is why delay and disconnect (faults of a connection) are not among the buttons.

Privacy: there is no backend. Nothing you type leaves your browser; the page sets no cookies and has no tracking. The server only delivers the static files (privacy notice).

First visit: the page downloads the Python runtime (several megabytes), which your browser then caches, so later visits start faster. It needs a browser with WebAssembly, which every current desktop and mobile browser has. If the simulator cannot start, the page shows a picture of the demo and the rest of the page still works. The demo is tested in Chromium and Firefox with Playwright on every change.

The same bench on your machine

pip install scpi-sim pyvisa pyvisa-py
curl -O https://raw.githubusercontent.com/steitec/scpi-sim/main/examples/profiles/bench.yaml
scpi-sim serve --bench bench.yaml

import pyvisa

rm = pyvisa.ResourceManager("@py")
opts = dict(read_termination="\n", write_termination="\n")
psu = rm.open_resource("TCPIP0::127.0.0.1::5025::SOCKET", **opts)
daq = rm.open_resource("TCPIP0::127.0.0.1::5026::SOCKET", **opts)
scope = rm.open_resource("TCPIP0::127.0.0.1::5027::SOCKET", **opts)
print(psu.query("*IDN?"))
psu.write("INST OUT1; VOLT 12; OUTP:SEL ON;:OUTP:GEN ON")
print(psu.query("MEAS:CURR?"))               # 0.500
print(daq.query("MEAS:VOLT:DC? (@101)"))     # about +1.2E+01
scope.write(":TIM:SCAL 0.0002")
print(scope.query(":MEAS:ITEM? FREQ,CHAN1"))  # about 1.0E+03

With scpi-sim serve --hislip --bench bench.yaml the same instruments also answer on TCPIP0::127.0.0.1::hislip_psu::INSTR — see VXI-11 and HiSLIP.

Running your own copy

The demo's source is in web/ and builds into a small nginx container:

git clone https://github.com/steitec/scpi-sim.git && cd scpi-sim
docker build -f web/Dockerfile -t scpi-sim-web .
docker run --rm -p 8080:80 scpi-sim-web     # http://localhost:8080

web/README.md describes the build (Pyodide is downloaded once and checked by SHA-256) and the tests.